Bluetooth: Conn: Bluetooth stack becomes unusable when communicating with both centrals and peripherals

[Ivan Herrera]

Follow up on: github.com/.../50438

Describe the bug
There is a specific case in which the Bluetooth stack seems to become stuck in a state where it will not accept/initiate any new connections. The following warning is printed in the console:
<wrn> bt_hci_core: opcode 0x2005 status 0x0c

I'm using an nRF52840 and NCS 2.1. The only way to get out of this state seems to be resetting the device. Below the steps I found which reproduces this state consistently using an example application.

To Reproduce

1. Grab two nRF52840 devkits and an iPhone that has nRFConnect for mobile installed.
2. Flash the "central_and_peripheral_hr" example on one devkit. Flash the "peripheral_hr" example on the other devkit.
3. Turn on the central_and_peripheral_hr devkit and connect to its console.
4. Get your phone and connect to "Nordic_HR". Make sure that the connection stays on for the following steps.
5. Turn on the "peripheral_hr" devkit. Make sure that it connects to the "central_and_peripheral_hr" devkit.
6. Once the connection has been established between the two devkits, turn off the "peripheral_hr" (important)
7. Close the connection with the phone through nRF Connect.
8. The Bluetooth stack is now in a state where it wont accept new connections (incoming/outgoing). Observe the console output in the section below.

Impact
This is not limited to the heartrate example. We initially discovered this bug in our own application and reproduced it using a sample so it would be easier to fix. Our users have to restart their gateways when configuring them to make sure that they still work properly. This feels like a pretty severe bug.

Logs and console output

*** Booting Zephyr OS build v3.1.99-ncs1-rc1  ***
Starting Bluetooth Central and Peripheral Heart Rate relay example
I: 8 Sectors of 4096 bytes
I: alloc wra: 1, fb8
I: data wra: 1, 164
I: SoftDevice Controller build revision: 
I: b3 30 26 8d c5 f3 4a bd |.0&...J.
I: f4 21 f5 83 2d de 9c f2 |.!..-...
I: 5e da 68 ec             |^.h.    
I: No ID address. App must call settings_load()
Scanning started
Advertising started
W: opcode 0x200a status 0x09
Connected: 48:79:82:C7:36:F1 (random).  <--- MOBILE PHONE
W: Ignoring data for unknown channel ID 0x003a
Filters matched. Address: F1:CE:C4:C1:92:BA (random) connectable: 1
Connected: F1:CE:C4:C1:92:BA (random) <--- HR PERIPHERAL
Security failed: F1:CE:C4:C1:92:BA (random) level 1 err 2
The discovery procedure succeeded
Heart Rate Sensor body location: Chest
Heart Rate Measurement notification received:

        Heart Rate Measurement Value Format: 8 - bit
        Sensor Contact detected: 1
        Sensor Contact supported: 1
        Energy Expended present: 0
        RR-Intervals present: 0

        Heart Rate Measurement Value: 92 bpm

Disconnected: F1:CE:C4:C1:92:BA (random) (reason 8) <--- HR PERIPHERAL
Disconnected: 48:79:82:C7:36:F1 (random) (reason 19). <--- MOBILE PHONE
Filters matched. Address: F1:CE:C4:C1:92:BA (random) connectable: 1 <--- HR PERIPHERAL
W: opcode 0x2005 status 0x0c
Connecting failed

Environment (please complete the following information):

• OS: MacOS
• Zephyr OS build v3.1.99-ncs1-rc1

[Hung Bui]

Hi Ivan, 

The case has been forwarded to the team. We will keep you updated. 

[Ivan Herrera]

Hi Hung! Thank you for your reply. Any updates on this, or foresight on how this will be fixed. Are we talking weeks or months? Just wondering so we can plan ahead for this.

[Hung Bui]

Hi Ivan, 

I'm afraid that I won't be able to tell when it will be fixed. I have seen some issues got a patch quickly in just a few days but some may take months. Sorry about that. 

[Ivan Herrera]

Ok thanks for letting me know! This doesn't seem like a minor issue so I hope it wont take months. Whenever you have two BLE devices interacting with a Nordic softdevice there's a chance to softbrick your device, which seems like a fairly common use-case for the nRF52. 

[Piotr]

Hello Ivan,

I've re-opened you original issue in GitHub. According to my investigation the issue is not caused by controller. Actually there are two issues that cause a behavior you have observed. Fix for one was already posted. For second one I need an advice from community. Please follow in GitHub and close this ticket.